MeSH terms and subject headings are the controlled vocabulary a database assigns to each record so that studies on the same topic are tagged consistently, no matter what words their authors used. MeSH is the Medical Subject Headings thesaurus used to index MEDLINE; other databases use their own equivalents, such as Emtree in Embase. Searching by heading retrieves records by their indexed concept rather than by surface wording, which is what gives a search much of its reach.
The problem subject headings solve
Authors describe the same idea in endlessly different ways. Free-text searching alone therefore misses any study that phrased the concept unexpectedly. A trained indexer reads each record and assigns the agreed heading for its topic, so one heading collects studies that used a dozen different phrasings. This is why a strong search strategy pairs headings with free-text terms: headings catch the inconsistently worded studies, free-text catches the newest records not yet indexed. Used together they are far more sensitive than either alone.
How a heading thesaurus is organised
Hierarchy and broader or narrower terms
Subject headings sit in a tree, with broad concepts at the top branching into narrower ones. Knowing where a term sits lets you choose the right level: too broad and you drown in irrelevant records, too narrow and you miss studies indexed under a sibling term. Browsing the hierarchy often reveals related headings you would never have guessed from keywords alone.
Explosion
Most databases let you explode a heading, which retrieves the term plus all of its narrower terms in one step. Exploding is usually what you want in a sensitive systematic search, because it captures the whole branch of a concept rather than a single node. You would only avoid it when a narrower term would clearly pull in irrelevant records.
Subheadings and qualifiers
Many headings can be paired with subheadings that narrow a concept to a specific aspect, such as a drug’s adverse effects or a disease’s therapy. Subheadings sharpen precision, but applied too aggressively they cut sensitivity, so reviewers use them sparingly in searches meant to find everything.
Why headings do not translate automatically
Each database maintains its own thesaurus, so a MeSH heading has no direct equivalent in Emtree, CINAHL headings, or other systems. Moving a strategy between databases for systematic reviews therefore means rebuilding the heading blocks by hand for each platform’s vocabulary, not pasting them across. This translation step is one of the more skilled parts of running a systematic literature search and a frequent source of error.
How indexers assign headings, and why recent records lag
A heading is not automatic. After a journal article is added to a database, a trained indexer reads it and assigns the headings that best describe its content, often marking one or two as the major topic of the paper with a star. That human step has two consequences for searching. First, indexing takes time, so the newest records may carry no headings yet or sit in an “in-process” status, which is exactly why a search must always pair headings with free-text terms to catch recent work. Second, indexing involves judgement, so two indexers can disagree and a study can end up under a heading you would not have predicted, which is part of why following citations between papers recovers studies that heading-based searching alone misses.
A worked heading block, step by step
Suppose the concept is depression. Building the heading side of the block in an Ovid MEDLINE interface looks like this:
- Look up the preferred heading in the thesaurus; the entry term “clinical depression” maps to the heading Depressive Disorder.
- Decide whether to explode it; here you would, to capture narrower headings such as major depressive disorder and dysthymia in one line, written
exp Depressive Disorder/. - Add the closely related heading
exp Depression/if the concept spans both the symptom and the disorder. - Join the headings to the free-text line, for instance
(depress* or dysthymi*).ti,ab., with OR to finish the concept block.
The judgement calls, whether to explode, whether to add a sibling heading, whether a subheading is safe, are where heading work rewards experience, and they map directly onto the eligibility rules set out in your inclusion and exclusion criteria.
Headings, free-text, and Boolean logic together
In a finished concept block, the exploded heading is joined to its free-text synonyms with OR, so a record matching either route is retrieved. The blocks are then combined with AND, using the rules in Boolean operators for literature searching. When the strategy is finalised, every heading and explosion is written into the search record under documenting and reporting your search and checked through PRESS peer review.
Common heading mistakes that lose studies
A few recurring errors with subject headings quietly undermine otherwise careful searches:
- Relying on headings alone. The newest records are not yet indexed, so a heading-only search misses recent studies; always add free-text synonyms.
- Forgetting to explode. Using a heading without exploding it drops every study indexed under a narrower term in the same branch.
- Over-applying subheadings. Restricting a heading to a subheading such as “therapy” cuts sensitivity, because indexers do not always attach the subheading you expect.
- Pasting headings between databases. A MeSH term has no automatic equivalent in Emtree or CINAHL headings, so each block must be rebuilt for that platform’s thesaurus.
Avoiding these is what lets controlled vocabulary do its job, deliver the sensitivity that turns a list of records into a defensible literature search the rest of the review can rely on.